MeitY Unveils DEPA Framework Handbook to Unlock Secure Data Collaboration for AI Training

Access to high-fidelity, representative training data has emerged as the defining bottleneck of the artificial intelligence boom, with enterprise data custodians routinely refusing to share datasets due to intellectual property theft risks, regulatory liabilities, and confidentiality concerns. To bridge this trust deficit, the Ministry of Electronics and Information Technology (MeitY) has officially unveiled the Data Empowerment and Protection Architecture (DEPA) Handbook for AI Training.

Unveiled by IT Secretary S Krishnan during an executive roundtable in Bengaluru, the framework was engineered in collaboration with think-tank iSPIRT Foundation. It establishes a standardized techno-legal blueprint that allows multi-party, cross-border research teams and commercial labs to collaborate on large-scale model training while ensuring participating institutions retain absolute sovereignty over their data.

The Architectural Core: 'Model Goes to Data' The foundational paradigm shift introduced by the DEPA training framework resolves traditional cybersecurity and IP leakage dilemmas:

• Inverted Logistics: Principal Scientific Adviser to the Government of India Prof. Ajay Kumar Sood emphasized that DEPA fundamentally flips traditional ingestion architecture. Rather than aggregating and transmitting vulnerable raw data into centralized compute environments, the AI training model travels into the data custodian’s secure perimeter.

• Controlled Computing Environments: Models execute compute cycles inside tamper-resistant, isolated enclaves where external parties cannot inspect, exfiltrate, or copy underlying datasets.

• Granular Data Provenance: The architecture incorporates immutable auditability, enabling data origins, specific usage constraints, and permission parameters to be tracked across the end-to-end model training lifecycle.

"India's AI ambitions must be matched by the ability to use data responsibly and securely. The DEPA Training Framework enables trusted data collaboration at scale, unlocking access to large-scale datasets for AI training while safeguarding privacy, commercial confidentiality and institutional control." — S Krishnan, Secretary, MeitY

India's Global Science Diplomacy Speaking on the geopolitical and multilateral dimensions of the initiative, Prof. Ajay Kumar Sood characterized DEPA as a cornerstone of India's international science diplomacy. By allowing universities, private healthcare networks, and sovereign institutions to engage in multi-party collaboration without relinquishing dataset control, the protocol aligns with global data governance initiatives championed by the United Nations and the World Bank.

Industry figures, including iSPIRT co-founder Sharad Sharma, have likened the milestone to the launch of the Unified Payments Interface (UPI), highlighting that standardizing trusted data exchange protocols at the infrastructure layer will be the decisive catalyst for developing inclusive, culturally representative, and resilient foundation models.